文献类型：期刊文献

中文题名：基于文本挖掘和机器学习的股指预测与决策研究

英文题名：Stock Forecast with Investors Sentiment by Text Mining and Machine Learning

作者：戴德宝[1];兰玉森[1];范体军[2];赵敏[3]

机构：[1]上海大学管理学院,上海200444;[2]华东理工大学商学院,上海200237;[3]上海大学悉尼工商学院,上海201800

年份：2019

卷号：0

期号：4

起止页码：166

中文期刊名：中国软科学

外文期刊名：China Soft Science

收录：CSTPCD;;北大核心:【北大核心2017】；CSSCI:【CSSCI2019_2020】；CSCD:【CSCD2019_2020】；

基金：国家自然科学基金重点项目(71431004);教育部人文社会科学研究规划基金项目(17YJA880014)

语种：中文

中文关键词：投资者情绪;股票预测;文本挖掘;机器学习

外文关键词：investors sentiment;stock forecast;text mining;machine learning

摘要：依据行为金融学理论,资本市场投资者的心理和行为对股票指数变动有重要影响。为此本文假设投资者情绪与股票指数存在一定内在作用机制,能预测股票市场整体价格变化。通过文本挖掘技术和情感分析方法生成积极和消极各三阶共六类投资者情绪时间序列数据;采用单位根检验、Granger因果关系检验和因子分析等方法构建上证投资者情绪综合指数,并分别使用支持向量机和神经网络预测股票市场价格变化,进行假设验证。结果表明:利用网络股市论坛文本数据和股票交易数据构建的上证投资者情绪综合指数能够提高股指走势预测的精度,有利于政府、在线平台、上市公司和投资主体更好决策。
According to the theory of behavioral finance,investors’psychological and behavior have important influence on the trend of stock market index.For this reason,this paper assumes that investors’sentiment is inherently associated with stock market index,which can predict the overall price change of a stock market.In this research,six kinds of the time series of investors sentiment are constructed by means of text mining technology and emotion analysis methods while Shanghai stock exchange composite investor sentiment index(SSECISI)is created by using unit root test,Granger causality test and factor analysis.The SVM and the neural network model are used to predict the stock market index to verify the correctness of the hypothesis.The results show that the SSECISI constructed by the forum mood and stock transaction data can improve the forecast precision of stock market index and enable government,online platforms,listed companies and investors to make decision better.